surface wind measurements in three gulf coast hurricanes ......surface wind measurements in three...

J. Wind Eng. Ind. Aerodyn. 98 (2010) 533–547

Contents lists available at ScienceDirect

Journal of Wind Engineeringand Industrial Aerodynamics

0167-61

doi:10.1

n Corr

E-m

(H.W. T

journal homepage: www.elsevier.com/locate/jweia

Surface wind measurements in three Gulf Coast hurricanes of 2005

Forrest J. Masters a,n, Henry W. Tieleman b, Juan A. Balderrama a

a Civil and Coastal Engineering, 365 Weil Hall, University of Florida, Gainesville, FL 32611, USAb Engineering Science and Mechanics, Virginia Polytechnic Institute and State University, 330 Norris Hall, Blacksburg, VA 24061, USA

a r t i c l e i n f o

Article history:

Received 3 June 2008

Received in revised form

20 April 2010

Accepted 23 April 2010Available online 11 June 2010

Keywords:

Tropical cyclone (hurricane)

Turbulence

Integral length scale

Gust factor

05/$ - see front matter Published by Elsevier

016/j.jweia.2010.04.003

esponding author. Tel.: +3 5 2 392 9537x150

ail addresses: [email protected] (F.J. Masters

ieleman), [email protected] (J.A. Balderrama).

a b s t r a c t

Detailed observations from surface layer experiments in the path of Atlantic hurricanes are presented in

this paper. The purpose is to obtain information that will aid in the reduction of hurricane wind damage

to residential structures by providing input for wind tunnel and full-scale flow simulation experiments

that assess wind loads on these structures under hurricane conditions. The contents of the paper

document the mean flow and turbulence characteristics from data recorded by nine mobile

instrumented towers deployed at coastal locations near the anticipated path of each of the three

hurricanes (Katrina, Rita, and Wilma) that made landfall on the US coast along the Gulf of Mexico in

2005. Wind data were collected at two elevations (5 and 10 m) from these towers by the Florida Coastal

Monitoring Program, a joint research program led by the University of Florida and the Institute for

Business and Home Safety.

Published by Elsevier Ltd.

1. Introduction

While numerous field experiments have been conductedin the last century to characterize wind in neutral conditions(e.g., Izumi, 1971), the literature is scarce in addressing surface-level winds occurring over land in tropical cyclones. Thisinformation is critical for many active areas in wind engineeringresearch, including active gust generation in wind tunnels (e.g.,Haan et al., 2006) and full-scale test facilities to evaluate buildingcomponents (e.g., Salzano et al., 2010). Recent analyses are largelyrestricted to open exposure measurements collected in a singlehurricane (e.g., Schroeder and Smith, 2003) or a few hurricanes(e.g., Yu et al., 2008). The purpose of this paper is to add to thisknowledge base through the analysis of data collected fromportable instrumented towers deployed by the Florida CoastalMonitoring Program (FCMP, fcmp.ce.ufl.edu) during three notablehurricanes in 2005: Katrina, Rita, and Wilma (Table 1). The FCMPis a research consortium – led by the University of Florida and theInstitute for Business and Home Safety – that focuses on full-scaleexperimental methods to quantify near-surface hurricane windbehavior and their resultant loads on residential structures. Since1998, the FCMP has collected over 50 observations in more than20 named storms in Alabama, Florida, Louisiana, Mississippi,North Carolina, and Texas.

Ltd.

5; fax: 3 5 2 392 3394.

), [email protected]

2. Instrumentation

The FCMP towers (shown in Fig. 1) are designed to resist a 90 m/s peak gust wind speed—this corresponds to a Category 5 hurricaneon the Saffir-Simpson Hurricane Wind Scale (www.nhc.noaa.gov/aboutsshws.shtml). Three levels of sensors outfit the tower atelevations of 3, 5, and 10 m. The data acquisition system measures3D wind speed and direction at the two upper levels and collectstemperature, rainfall, barometric pressure, and relative humiditydata at the tower’s base. Two RM Young anemometer systems – awind monitor (Model no. 05103 V) and a custom array of three gillpropellers (Model no. 27106 R) – collect data at the 10 m level.A second array of gill propellers collects wind speed data at the 5 mlevel to measure winds at the approximate mean roof height of asingle-story home. Dynamic characteristics of the anemometer’sfour-blade polypropylene helicoid propellers (Model no. 08234)include a 2.7 m 63% recovery distance constant and a dampednatural wavelength of 7.4 m. The wind monitor is rated for a 100 m/sgust survival and has a 50% recovery vane delay distance of 1.3 m.The limitations caused by its frequency response characteristics aredetailed in Schroeder and Smith (2003). The wind monitor serves asa redundant system by providing a second set of 10 m wind velocitydata that is used to quality control the data recorded by the primarysystem (the array of three gill propellers).

The system described is capable of making velocity observa-tions on a continuous basis with a sampling frequency of 10 Hz.Data observations are recorded, post-processed, uploaded inreal-time, and made available from the FMCP web site (http://fcmp.ce.ufl.edu) within minutes, providing mean wind andturbulence characteristics of the surface flow without interrup-tions for at least 24 h.

http://fcmp.ce.ufl.edu

http://fcmp.ce.ufl.edu

www.elsevier.com/locate/jweia

dx.doi.org/10.1016/j.jweia.2010.04.003

mailto:[email protected]




dx.doi.org/10.1016/j.jweia.2010.04.003

Table 1Hurricanes of 2005: Mobile towers and periods of data acquisition (UTC).

Hurricane Tower # of files Data acquisition Passage of eye or velocity peak

Begin End Begin End

Katrina T0 92 28-Aug-05 15:04 29-Aug-05 13:49 29-Aug-05 12:00 29-Aug-05 14:00

T1 104 28-Aug-05 21:43 29-Aug-05 23:28 29-Aug-05 11:00 29-Aug-05 12:00

T2 56 29-Aug-05 02:37 29-Aug-05 16:22 29-Aug-05 11:00 29-Aug-05 12:00

Rita T0 92 23-Sep-05 16:54 24-Sep-05 15:39 24-Sep-05 08:00 24-Sep-05 09:00

T3 74 24-Sep-05 02:00 24-Sep-05 20:15 24-Sep-05 08:00 24-Sep-05 11:00

T5 102 23-Sep-05 21:00 24-Sep-05 22:15 24-Sep-05 08:00 24-Sep-05 10:00

Wilma T0 75 23-Oct-05 22:55 24-Oct-05 17:25 24-Oct-05 10:00 24-Oct-05 12:00

T1 57 24-Oct-05 07:54 24-Oct-05 21:54 24-Oct-05 11:30 24-Oct-05 14:30

T2 62 24-Oct-05 01:19 24-Oct-05 16:34 24-Oct-05 11:00 24-Oct-05 13:00

Fig. 1. Florida Coastal Monitoring Program tower platform shown in the

foreground. Stowed tower shown in the background.

F.J. Masters et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 533–547534

3. Data analysis

The gill anemometers are mounted on the tower such that thebisector of the solid angle formed by the three anemometers ishorizontal. Consequently, the obtained data must be transformedfrom the anemometer-oriented coordinate system to the meanwind coordinate system with the u and v components in thehorizontal plane and the w component vertically upward. Thex-coordinate is oriented towards the direction of the mean flow,the y-coordinate is perpendicular to the x-direction in thehorizontal plane, and the z-coordinate is directed verticallyupward to form a right-handed coordinate system. This transfor-mation must be performed for each of the simultaneousobservations recorded by the system of anemometers at eachlevel and repeated at each time step to obtain a continuous timehistory for each sample record.

The discussion in this paper is based on results obtained for the15-min time histories of each sample record, including the meanvelocity U and its direction, and the three turbulence intensitycomponents sa/U for a¼u, v, and w, respectively. Also obtainedwere the three covariances uw, uv, and vw, the turbulence integralscales for the three velocity components, and the maximum 3 sgust for each sample record.

4. Tower locations

Only observations from those towers that were the closest tothe path of the respective hurricanes were considered for analysisin this paper. Tower locations are shown in Fig. 2 and their GPScoordinates are provided in Table 2. The perpendicular distancefrom the tower location to the path of the hurricane ranged from3.2 km (Rita, tower T3) to 74 km (Katrina, tower T2) with a meandistance of 28.8 km for all nine deployments. For a tower locateddirectly in the path of the hurricane eye, one can expect airvelocity to increase, then stall during a calm period, andsubsequently, increase abruptly as the eye passes the towerlocation. For a tower located outside the eye at all times, thevelocity exhibits just a single peak. Katrina T1, and Rita T0experienced single peaks. The remaining towers employed for thisresearch were subjected to the passage of a hurricane eye; thus,they sustained a double velocity peak.

In general, the towers were deployed in open terrain, mostlyareas with uniform roughness within a radial distance of 400 mfrom the tower location. The exceptions are towers T3 and T5 inHurricane Rita; the former was located downwind of a subdivi-sion when the winds traveled from the north, while the latter wassituated 100 m from tall trees when the winds traveled from thenorth and west. The average distance from the coast, measuredparallel to the path of the hurricane, for the nine towers (three foreach hurricane) was 41.8 km.

5. Mean wind speeds

The passage of the eye of a hurricane through a tower locationis characterized by increasing wind speeds followed by arelatively short period of calm winds, after which the windvelocity increases again to a second maximum before graduallydecreasing as the hurricane moves away from the tower. Thisvariation of the mean wind distribution over time is depicted inFigs. 3 and 4 for Wilma tower T0 and Rita tower T5, respectively.Some of the observed wind distributions, Katrina towers T0 andT1 (Fig. 5) and Rita tower T0, do not exhibit this typical windpattern, instead they show a single maximum. The observations ofthe Katrina T0 tower were abruptly terminated due to a powerfailure and did not record the maximum speed. The winddirection change of 1801 during the passage of the eye or afterthe single peak wind speeds was observed in all other cases. Fortowers located on the starboard side of the hurricane path, themean wind rotates in a clockwise fashion; while for those locatedon its port side, the mean wind rotates in the opposite direction asthe eye or the peak passes (Table 3).

Fig. 2. Location of FCMP towers during Hurricanes: (a) Katrina, (b) Rita, and

(c) Wilma.

Table 2GPS coordinates of the mobile towers.

Hurricane Tower Latitude Longitude

Katrina T0 30.3801 �89.4551

T1 29.8253 �90.0319

T2 29.4441 �90.2628

Rita T0 29.9512 �94.0220

T3 29.9548 �93.9542

T5 30.0797 �93.7841

Wilma T0 25.9008 �81.3114

T1 26.1458 �80.5067

T2 25.8681 �80.8997

F.J. Masters et al. / J. Wind Eng. Ind. Aerodyn. 98 (2010) 533–547 535

The maximum observed 15-min wind speeds at the 10 m levelvaried between 23.0 and 36.0 m/s, while the maximum 60 swind speeds varied between 29.2 and 41.9 m/s. The latter

indicates that adjusting for terrain, the intensity of the hurricanesat the tower locations fell in the category 1 or below according tothe Saffir-Simpson Hurricane Wind Scale (www.nhc.noaa.gov/aboutsshws.shtml).

6. Turbulence intensities

Near surface turbulence is caused by the earth’s landscape,which acts as a momentum sink (Wieringa, 1993). It has beensuggested that organized storm features aloft can contribute to orpossibly modulate surface level wind characteristics owing totheir convective nature (Bradbury et al., 1994; Wurman andWinslow, 1998). Based on 15-min turbulence intensities for eachof the towers that experienced an eyewall passage, it appears thatthe turbulence intensities were not affected by the passage of thehurricane eye. Instead, the turbulence intensities depend primar-ily on the roughness of the upwind terrain.

The records were partitioned into two sections for which thewind direction range was approximately 301; Katrina T0 wasthe exception, producing only one section. Average values forturbulence intensities at 5 and 10 m for each partition, along withtheir corresponding wind direction ranges, are shown in Table 4.Partitions (a) and (b) for Katrina T1 (Fig. 6), Katrina T2, Rita T0,and Wilma T1 and partition (b) for Rita T3 and T5(Group 1) exhibit average turbulence intensities at 10 m,su/U¼16.8%, sv/U¼12.4%, and sw/U¼6.8%, corresponding torelatively smooth terrain; these results are similar to the 5 minturbulence intensities (su/U¼17.6%, sv/U¼16.4%, and sw/U¼8.5%)obtained by Schroeder and Smith (2003) for an airport exposureduring Hurricane Bonnie of 1998. On the other hand, partitions (a)and (b) for Wilma T0 and T2, partition (a) for Rita T3 and T5(Fig. 7), and the partition for Katrina T0 (Group 2) produced highervalues for the averaged turbulence intensities at 10 m,sa/U¼23.2%, 17.8%, and 10.4% for a¼u, v, and w respectively;Schroeder et al. (2009) performed a similar analysis from 10.7 mdata for seven tropical cyclones and obtained a comparableaverage longitudinal turbulence intensity value, 23.8%, for roughterrain (9.00 cmoz0 o18.99 cm). The lower values of su/U andsv/U obtained for the first group of partitions are associated withterrain whose aerodynamic roughness length, z0, is of the order of1–3 cm. The higher turbulence intensities observations (Group 2)have corresponding z0 values of 13.3 and 7.2 cm for su/U and sv/U,respectively.

Observed averages of su/sv and su/sw are 1.34 and 2.38,respectively. The limited frequency response of the propelleranemometers may lead to partial filtering of the higher frequenciesof the vertical velocity component, thus resulting in valuessystematically larger than the customary value of 2.0 for su/sw.Plots of u and w spectra reveal that the vertical-velocity spectrum isshifted towards higher frequencies with respect to the u-spectrum.

Fig. 3. Mean wind and direction observations from tower T0, Hurricane Wilma (UTC).

Fig. 4. Mean wind and direction observations from tower T5, Hurricane Rita (UTC).


Consequently, the limited frequency response of the anemometerswill affect sw more than su, and the ratio su/sw is expected to belarger than normal value of 1.92 (Panofsky and Dutton, 1984).

Some of the turbulence intensities, especially those from theRita towers, reveal a decrease in magnitude with increasing meanvelocity before the eye passes the tower location. For example, the

Fig. 5. Mean wind and direction observations from tower T1, Hurricane Katrina (UTC).

Table 3Maximum velocities and wind direction rotation.

Hurricane

and tower

Height, m Velocity

peaks

Change in

wind direction, degrees

Maximum mean velocity (m/s) Maximum 3-s gust (m/s)

First peak Second peak First peak Second peak

Katrina T0 5 – +60 24.1 – 37.9 –

10 – +60 26.0 – 40.6 –

Katrina T1 5 Single �175 25.3 – 40.5 –

10 Single �160 30.5 – 45.2 –

Katrina T2 5 Double �130 25.6 25.0 40.3 36.6

10 Double �130 28.3 28.9 42.3 41.9

Rita T0 5 Single �150 33.0 – 47.4 –

10 Single �150 36.0 – 50.0 –

Rita T3 5 Double �140 20.7 14.7 35.7 24.5

10 Double �150 23.8 18.3 39.8 27.8

Rita T5 5 Double +200 27.2 20.9 42.4 28.1

10 Double +200 27.5 22.0 42.4 29.0

Wilma T0 5 Double +185 23.0 23.1 37.4 39.0

10 Double +185 25.6 26.3 38.0 42.5

Wilma T1 5 Double +160 25.4 28.0 42.4 39.2

10 Double +150 34.3 35.3 45.6 44.9

Wilma T2 5 Double +160 27.8 22.3 40.1 41.9

10 Double +160 29.7 27.0 41.6 48.4

Note: Katrina T0 tower experienced a power failure and did not record the maximum wind speed.


10 m longitudinal turbulence intensity at tower T5 decreasedapproximately (25–20%)/25%¼20% near the approach of theeyewall (Fig. 7). This trend occurred because of experimentaldesign and is not related to the storm’s passage. The towers wereintentionally oriented so that the maximum winds occurred in thedirection corresponding to the least built-up terrain.

7. Friction velocity, Un

Covariances of the fluctuating velocity components wereobtained from the time histories of the u, v, and w componentsof each 15 min sample record. It was found that the uw covariancehas mostly negative values, as expected in a standard boundary

Table 4Turbulence intensities with standard deviations (%).

Hurricane and towersu=U7StD sv=U7StD sw=U7StD

Mean wind

direction (deg)

Wind direction

range (deg)Segment range

Observations at z¼5 m

Katrina T0 25.7274.14 22.8472.05 8.2971.24 63 43–80 75–92

Katrina T1 (a) 17.3573.52 14.2275.51 11.8271.47 33 16–47 9–58

Katrina T1 (b) 24.2974.47 16.0972.69 9.1470.99 244 226–276 70–104

Katrina T2 (a) 18.2971.79 14.9772.16 6.4770.52 24 5–38 1–31

Katrina T2 (b) 17.6971.03 12.8570.87 6.1571.12 273 261–295 46–56

Rita T0 (a) 17.8271.44 13.8572.01 4.9470.62 6 �11–21 1–61

Rita T0 (b) 16.0270.98 12.4070.88 5.8770.67 230 220–247 72–92

Rita T3 (a) 32.4672.15 24.2172.85 15.5771.90 358 341–366 1–27

Rita T3 (b) 21.2771.63 12.9471.15 10.6771.55 224 218–237 33–71

Rita T5 (a) 24.8872.51 20.4371.87 8.5271.08 16 7–33 1–45

Rita T5 (b) 17.0671.68 11.4970.78 5.4070.30 211 192–227 55–91

Wilma T0 (a) 22.9675.06 16.4771.58 8.6872.41 117 109–129 1–46

Wilma T0 (b) 23.8372.98 16.5971.09 10.0070.61 301 295–314 54–75

Wilma T1 (a) 23.6873.26 13.1672.50 7.6471.43 126 117–139 1–17

Wilma T1 (b) 17.1171.30 10.3972.70 7.4471.12 277 258–294 25–57

Wilma T2 (a) 23.5572.24 16.3972.35 11.0772.09 136 121–165 1–40

Wilma T2 (b) 22.0872.06 16.0871.60 12.9371.72 287 269–296 47–62


Katrina T0 24.8873.75 21.1472.00 10.3970.56 67 47–83 75–92

Katrina T1 (a) 17.1472.69 14.3774.15 7.0870.29 32 16–49 9–58

Katrina T1 (b) 16.7371.72 12.8771.27 7.0070.44 248 235–270 72–104

Katrina T2 (a) 16.8471.87 12.5372.19 7.9970.51 23 8–37 1–31

Katrina T2 (b) 15.9171.50 11.9770.76 6.7070.84 272 260–294 46–56

Rita T0 (a) 16.9871.23 13.0272.05 5.7470.70 7 �10–23 1–61

Rita T0 (b) 15.4570.92 12.3670.60 6.0570.41 231 221–249 72–92

Rita T3 (a) 29.0671.85 19.0671.40 12.1871.37 3 �15–10 1–27

Rita T3 (b) 18.6971.88 12.6871.47 8.7171.17 226 219–245 33–71

Rita T5 (a) 24.3572.38 21.2271.80 10.8071.51 16 5–36 1–45

Rita T5 (b) 16.7071.50 11.7970.62 6.4870.30 210 190–227 55–91

Wilma T0 (a) 20.1674.65 13.8671.70 9.3273.51 117 110–128 1–46

Wilma T0 (b) 20.7573.28 17.2471.21 10.3470.83 299 293–313 54–75

Wilma T1 (a) 18.1371.91 11.2271.30 6.4670.93 132 125––143 1–17

Wilma T1 (b) 15.7271.74 11.4172.32 6.2170.78 284 270–295 25–57

Wilma T2 (a) 21.5072.31 16.2771.96 9.6372.07 133 120–161 1–40

Wilma T2 (b) 21.4171.70 15.9371.79 9.9371.10 285 269–296 47–62

Fig. 6. 15 min, 10 m turbulence intensity observations from tower T1 during Hurricane Katrina (passage of the eye between 11:00 and 12:00 UTC).


Fig. 7. 15 min, 10 m turbulence intensity observations from tower T5 during Hurricane Rita (passage of the eye between 08:00 and 10:00 UTC).

Table 5Distribution of positive and negative uw covariances.

Hurricane and tower # of files Height (m) # of files with uw covariances

Negative values Positive values

Katrina T0 92 5 22 70

10 90 2

Katrina T1 104 5 45 59

10 104 0

Katrina T2 56 5 53 3

10 49 7

Rita T0 92 5 80 12

10 92 0

Rita T3 74 5 74 0

10 74 0

Rita T5 102 5 40 62

10 0 102

Wilma T0 75 5 75 0

10 54 21

Wilma T1 57 5 57 0

10 51 6

Wilma T2 62 5 59 3

10 58 4


layer under thermally neutral conditions (Table 5). The presenceof positive values indicates that the turbulent momentum flux isaway from the boundary rather than towards it.

For the Rita T5 tower, which had tall trees to the northand west at a distance of about 100 m, the uw covariance atthe 10-m level was entirely positive and 61% of its 5-m levelvalues were positive. For the remaining eight towers, the meanpercentage of negative averaged uw covariance was 80% atthe 10 m level and 65% at the 5 m level. The friction velocityfor each sample record was obtained from the quartic root ofthe resultant of the two horizontal turbulent stresses, uw and vw

(Weber, 1999). These results were then used to obtain theturbulence ratios, sa/Un, and to estimate values for the roughnesslength, z0.

The averaged turbulence ratios from the hurricane observa-tions are sa/Un

¼3.09, 2.36, and 1.29 for a¼u, v, and w,respectively (Table 6). These results compare reasonably wellwith the flat, uniform, and smooth multi-terrain observationsreported by Hogstrom (1990), for which sa/Un

¼2.78, 2.44, and1.24 for a¼u, v, and w, respectively.

8. Aerodynamic roughness length, z0

Since roughness lengths based on mean velocities at twolevels must not be used, one must resort to turbulence observa-tions to obtain more reliable values for the roughness lengths.These lengths are evaluated using the hurricane data and Eqs. (1)

Table 6Nondimensional turbulence ratios.

Hurricane and tower su=U�7StD sv=U�7StD sw=U�7StD

Observations at z¼5 mKatrina T0 4.8371.16 4.3971.30 1.6070.53

Katrina T1 (a) 2.1570.49 1.7170.52 1.4970.41

Katrina T1 (b) 2.9970.41 1.9970.35 1.1470.20

Katrina T2 (a) 3.2770.53 2.6570.41 1.1670.17

Katrina T2 (b) 3.5570.22 2.5870.27 1.2270.15

Rita T0 (a) 3.7271.11 2.8770.80 1.0170.21

Rita T0 (b) 2.6370.48 2.0570.43 0.9570.07

Rita T3 (a) 3.1770.32 2.3570.26 1.5170.14

Rita T3 (b) 3.2270.74 1.9470.37 1.6370.51

Rita T5 (a) 2.8470.34 2.3570.42 0.9770.09

Rita T5 (b) 5.9271.46 4.0271.08 1.9070.54

Wilma T0 (a) 3.2970.81 2.4370.72 1.2470.29

Wilma T0 (b) 2.6070.24 1.8270.22 1.0970.08

Wilma T1 (a) 4.3171.55 2.3470.66 1.3670.42

Wilma T1 (b) 2.5170.50 1.5270.43 1.0770.14

Wilma T2 (a) 2.6570.62 1.8770.58 1.2070.10

Wilma T2 (b) 3.1170.91 2.2570.62 1.8370.60


Katrina T0 4.6971.05 4.0271.02 1.9770.44

Katrina T1 (a) 3.0070.48 2.5370.80 1.2470.13

Katrina T1 (b) 2.4170.27 1.8570.20 1.0170.07

Katrina T2 (a) 2.8570.34 2.1270.39 1.3670.14

Katrina T2 (b) 4.3271.81 3.3471.69 1.7970.68

Rita T0 (a) 3.8570.73 2.9670.73 1.3070.26

Rita T0 (b) 3.8871.67 3.0971.28 1.5070.59

Rita T3 (a) 2.3270.28 1.5270.20 0.9670.06

Rita T3 (b) 2.2770.18 1.5470.11 1.0670.08

Rita T5 (a) 2.5670.40 2.2470.40 1.1370.13

Rita T5 (b) 4.1970.66 2.9770.52 1.6370.24

Wilma T0 (a) 2.1370.31 1.4970.24 0.9770.10

Wilma T0 (b) 2.8470.70 2.3670.45 1.4070.21

Wilma T1 (a) 2.4070.18 1.4970.18 0.8670.10

Wilma T1 (b) 3.8271.63 2.8271.38 1.5070.59

Wilma T2 (a) 2.8571.02 2.1870.95 1.2170.25

Wilma T2 (b) 2.1770.34 1.6170.22 1.0070.11

Table 7Aerodynamic roughness lengths.

Hurricane

and tower

Height (m) (z0)1 (cm) (z0)2 (cm) Mean wind

direction (deg)

(z0)3 (cm) Mean wind

direction (deg)


direction (deg)

Katrina T0 5 0.2 0.5 63 0.270.1 60 – –

5 – – – 0.370.2 33 – –

10 0.5 0.7 67 1.971.0 62 – –

10 – – – 1.870.9 37 – –

Katrina T1 5 3.2 3.9 33 7.171.7 36 4.972.7 37

5 3.3 2.9 244 2.170.6 237 9.373.5 242

10 0.9 0.9 32 1.270.3 28 – –

10 3.1 3.6 248 1.270.6 248 3.371.5 246

Katrina T2 5 0.4 0.6 24 0.370.1 25 0.470.3 28

5 0.2 0.2 273 0.170.1 301 – –

10 1.1 1.1 23 0.670.3 23 1.671.0 24

10 0.01 – 272 0.0470.04 300 – –

Rita T0 5 0.1 0.3 6 0.270.2 3 – –

5 0.7 – 230 0.470.2 227 – –

10 0.1 0.2 7 0.170.0 4 – –

10 0.05 – 231 0.270.1 229 – –

Rita T3 5 9.9 9.3 358 11.772.0 359 12.773.9 359

5 1.2 1.6 224 0.870.3 221 2.070.6 222

10 40.4 40.6 3 35.2712.7 3 56.3713.4 5

10 7.4 8.5 226 5.871.3 222 5.571.6 222

Rita T5 5 5.0 4.5 16 1.070.4 12 6.572.8 15

5 – – 211 0.0170.01 210 – –

10 14.2 13.9 16 9.172.0 9 21.278.1 13

10 0.01 – 210 0.270.1 210 – –

Wilma T0 5 1.4 2.3 117 2.070.9 117 – –

5 6.1 5.8 301 4.871.4 301 6.170.8 298

10 13.3 14.5 117 6.572.6 117 17.074.1 117


Table 7 (continued )

Hurricane

and tower

Height (m) (z0)1 (cm) (z0)2 (cm) Mean wind

direction (deg)


direction (deg)


direction (deg)

10 4.1 0.2 299 14.674.3 300 11.472.5 117

Wilma T1 5 0.3 0.4 126 1.771.4 126 2.972.5 121

5 1.4 2.3 277 0.470.2 288 0.670.3 288

10 4.7 4.4 132 0.770.3 129 5.772.2 131

10 0.04 0.07 284 0.470.3 289 1.370.7 289

Wilma T2 5 – – – 7.571.6 132 10.874.5 130

5 5.2 4.6 136 2.871.2 137 – –

5 1.9 3.3 287 8.272.1 289 – –

10 4.2 4.0 133 16.874.2 127 19.676.6 128

10 17.2 22.4 285 10.975.2 287 20.1713.9 284

Note 1: ‘–’ denotes insufficient data to perform calculations.

Note 2: (z0)1 and (z0)2 were calculated for each partition, while (z0)3 and (z0)4 were calculated from subsets of each partition.

Table 8Turbulence integral scales with standard deviations (m).

Hurricane and tower Lux 7StD (m) Lv

x 7StD (m) Lwx 7StD (m)


Katrina T0 127.8743.4 102.1744.8 18.175.7

Katrina T1 (a) 225.67109.5 155.17190.0 49.7728.5

Katrina T1 (b) 83.2734.9 84.4732.2 18.276.3

Katrina T2 (a) 146.9791.3 98.6795.2 18.178.9

Katrina T2 (b) 98.3727.4 87.3746.2 12.274.8

Rita T0 (a) 135.4747.3 115.1777.2 20.376.2

Rita T0 (b) 89.3729.7 66.4734.3 14.172.8

Rita T3 (a) 47.2721.1 34.4720.5 5.771.4

Rita T3 (b) 80.4728.6 29.0711.6 13.673.6

Rita T5 (a) 104.5751.1 66.1755.3 15.973.2

Rita T5 (b) 136.5751.4 92.2764.6 18.174.8

Wilma T0 (a) 95.8738.0 57.5727.8 16.1715.8

Wilma T0 (b) 101.0733.8 55.4719.9 9.571.7

Wilma T1 (a) 97.4762.8 66.6732.9 26.7751.8

Wilma T1 (b) 136.3790.9 158.37173.6 24.1733.2

Wilma T2 (a) 88.6737.7 70.7737.1 10.972.8

Wilma T2 (b) 146.4753.8 59.2740.4 18.575.9


Katrina T0 133.2738.9 110.7754.0 18.975.6

Katrina T1 (a) 267.47165.7 243.57264.2 22.078.8

Katrina T1 (b) 135.1751.2 68.6719.9 20.374.2

Katrina T2 (a) 170.87110.5 121.37108.3 19.974.4

Katrina T2 (b) 126.2743.0 104.5749.8 22.0711.8

Rita T0 (a) 147.2750.4 126.4784.9 20.175.3

Rita T0 (b) 101.0735.6 64.8739.2 15.172.9

Rita T3 (a) 74.4731.9 55.1749.1 11.273.3

Rita T3 (b) 112.8741.9 49.4721.5 16.872.8

Rita T5 (a) 107.7749.7 65.5752.1 18.874.2

Rita T5 (b) 150.4758.3 84.7759.7 19.175.1

Wilma T0 (a) 128.5749.7 62.8726.4 23.7722.5

Wilma T0 (b) 152.9760.4 66.0739.7 25.579.5

Wilma T1 (a) 160.6778.4 86.2741.9 24.1711.5

Wilma T1 (b) 189.17142.2 106.3763.0 13.173.5

Wilma T2 (a) 120.4754.4 59.5730.5 16.574.0

Wilma T2 (b) 173.5764.4 85.3749.9 25.678.8


and (2)—expressions derived from the logarithmic velocity law,where the avg subscript refers to the average values and the ind

subscript refers to the respective individual values.

z0 ¼ elnðzÞ�0:4ðU=U�Þ ð1Þ

z0 ¼ elnðzÞ�0:4ðððsw=U�Þavg Þ=ððsw=UÞindÞÞ ð2Þ

The approaches taken to derive the roughness lengths are asfollows:

1.
(z0)1 is derived by inserting the average value of U/Un for eachpartition into Eq. (1).
2.
(z0)2 is derived by performing a linear regression analysis on U
as a function of Un with least squares criterion (assuming they-interceptE0) and evaluating the slope of the plot of U versusUn for each partition to insert it into Eq. (1).

3.
(z0)3 is derived from Eq. (2) using the average value of sw/Un foreach data set and the individual values of sw/U in order to obtaina set of roughness lengths that are subsequently averaged. Thisoperation is conducted for ranges in the data set where theobservations of sw/U are reasonably uniform and where thechange in wind direction is within an acceptable range.
4.
(z0)4 is derived from Eq. (1) using individual values of theparameter U/Un to obtain a set of roughness lengths that can be


subsequently averaged for ranges where U/Un is reasonablyuniform and where the change in wind direction is within atolerable range.

In most cases, the spread in the z0 values for a single height/tower combination is confined to one decade. Ideally, the 5 m andthe 10 m roughness lengths should be the same or at least verysimilar. This similarity of the 5 and 10 m roughness length exists insome approaches for Katrina T1, Katrina T2, Rita T0, and Wilma T1(Table 7). However, for the remaining towers, the 10 m roughnesslength is larger than that obtained from data at 5 m. This may be

Fig. 8. Distribution of the Lux integral scale for hurricane Katrina, tower T1

Fig. 9. Distribution of the Lvx integral scale for hurricane Katrina, tower T1

attributed to the 5 m measurement being more sensitive to theimmediate upwind fetch, which is usually flat, open terrain, thanthe 10 m observation. The 10 m measurement is more sensitive tothe larger roughness obstacles (such as large vegetation orbuildings) further upstream from the observation site.

9. Turbulence integral length scales

The three turbulence integral length scales, Lux , Lv

x , and Lwx , were

determined for each 15 min sample record, their average values

at z¼5 and 10 m (passage of the eye between 11:00 and 12:00 UTC).

at z¼5 and 10 m (passage of the eye between 11:00 and 12:00 UTC).

Table 11Average 3-s gust and peak factors.

Hurricane and tower GF(3 s, z) g(3 s, z)

Observations at z¼5 mKatrina T0 1.841 3.246

Katrina T1 (a) 1.617 3.579

Katrina T1 (b) 1.744 3.037

Katrina T2 (a) 1.581 3.168

Katrina T2 (b) 1.523 2.947

Rita T0 (a) 1.569 3.190

Rita T0 (b) 1.489 3.058

Rita T3 (a) 2.028 3.152

Rita T3 (b) 1.664 3.114

Rita T5 (a) 1.819 3.279


are shown in Table 8 for each partition. Several differentapproaches have been employed in the literature to determinethe length scale. For this analysis, sequential segments werelinearly detrended and their along-wind velocity componentswere calculated. The scaled covariance (zero-mean and unitvariance autocorrelation) function was then computed throughWiener–Khinchine relations, specifically through an inverseFourier transform of the autospectrum estimate. Finally, thescaled covariance function was integrated numerically fromt¼0 s to the first crossing of the time lag, t-axis, and multipliedby the segment’s mean wind speed to estimate the length scale.

The observed values generally compare favorably with ob-servations made by Counihan (1975), who estimated Lu

x¼196 mfor z0¼0.01 m and Lu

x¼139 m for z0¼0.03 m. However, the peakintegral length scales from each tower frequently exhibited largevalues (as large as 1200 m) for both horizontal scales (Figs. 8 and9 for Katrina T1). The maximum values for the vertical scale, Lw

x ,were in the 50 m range. The correlation between each of thehorizontal scales at the 5 m and 10 m levels is quite high, most ofits correlation coefficients exceed 0.9 (Table 9).

The correlation coefficients of the vertical scales at the twolevels (r values below 0.5) are much lower than those of thehorizontal scales. The results also show a low correlation betweenthe Lu

x integral scale and the 3 s gust velocity (correlationcoefficients well below 0.5 are shown in Table 10); as a result,large values of the integral scales cannot be associated with highvalues of the short duration peak velocities. Thus, it would not beexpected that these large eddies are responsible for isolatedswaths of damage to buildings and trees observed during damageassessments (e.g., Wakimoto and Black, 1994).

Rita T5 (b) 1.513 3.012

Wilma T0 (a) 1.732 3.217

Wilma T0 (b) 1.744 3.126

Wilma T1 (a) 1.685 2.880

Wilma T1 (b) 1.492 2.874

Wilma T2 (a) 1.751 3.190

Wilma T2 (b) 1.735 3.330


Katrina T0 1.798 3.180

Katrina T1 (a) 1.561 3.267

10. 3-s gust velocities

In addition to the 15 min mean velocity, the database alsoincludes the maximum 3 s gust velocity for each 15 min samplerecord. The latter is the basic wind speed used in the ASCE 7-10

Minimum Design Loads for Buildings and other Structures (ASCE 7-10,

Table 9Correlation coefficients between Lu

x , Lvx and Lw

x at 5 and 10 m.

Hurricane and tower Lux Lv

x Lwx

Katrina T0 0.944 0.950 0.406

Katrina T1 0.790 0.884 0.347

Katrina T2 0.927 0.957 0.387

Rita T0 0.917 0.956 0.487

Rita T3 0.827 0.804 0.694

Rita T5 0.967 0.984 0.498

Wilma T0 0.792 0.989 0.242

Wilma T1 0.899 0.886 �0.096

Wilma T2 0.900 0.921 0.414

Table 10Comparison of Lu

x and U(3 s) at 10 m.

Hurricane and tower Mean Lux (m) Mean U(3 s) (m/s) Standard dev

Katrina T0 95.5 13.94 48.3

Katrina T1 204.0 24.50 136.7

Katrina T2 156.1 30.53 87.5

Rita T0 140.7 25.66 56.4

Rita T3 95.7 24.74 41.0

Rita T5 129.5 23.17 67.8

Wilma T0 136.6 19.94 53.6

Wilma T1 176.4 27.23 117.9

Wilma T2 138.8 22.90 65.0

2010) at 10 m above the ground for a category C roughnessexposure (Zhou and Kareem, 2002). The 3 s gust factor at a heightof z meters above the ground, G(3 s, z), and the normalized gust orthe peak factor, g(3 s, z), are defined in Eqs. (3) and (4),respectively.

Gð3s, zÞ ¼Uð3s, zÞ

UðzÞð3Þ

gð3s, zÞ ¼Uð3s, zÞ�UðzÞ

su¼

Uð3s, zÞ

su�TIu ð4Þ

The 3 s gust speed, U(3 s, z) from Eq. (4), can be substituted intoEq. (3) and one can obtain an expression for the 3 s gust factor,

iation Lux (m) Standard deviation U(3 s) (m/s) Correlation coefficient

8.81 0.550

8.66 0.026

6.37 0.128

7.9 0.292

6.16 �0.015

6.30 0.347

8.54 0.269

9.79 �0.038

9.20 0.501

Katrina T1 (b) 1.530 3.158

Katrina T2 (a) 1.542 3.213

Katrina T2 (b) 1.504 3.149

Rita T0 (a) 1.543 3.198

Rita T0 (b) 1.469 3.032

Rita T3 (a) 1.812 2.795

Rita T3 (b) 1.572 3.068

Rita T5 (a) 1.795 3.253

Rita T5 (b) 1.503 3.016

Wilma T0 (a) 1.620 3.089

Wilma T0 (b) 1.626 3.004

Wilma T1 (a) 1.551 3.029

Wilma T1 (b) 1.463 2.936

Wilma T2 (a) 1.662 3.079

Wilma T2 (b) 1.664 3.102


G(3 s, z), as a function of the turbulence intensity (Eq. (5)).

Gð3s, zÞ ¼ 1þgð3s, zÞ

UðzÞsu

¼ 1þgð3s, zÞTIu ð5Þ

Substituting the log law for neutral atmospheric conditions andaccepting that su/Un and von Karman’s constant are 2.5 and 0.4,respectively, Eq. (5) for the gust factor, G(3 s,z), can be rewrittenas function of roughness length (Eq. (6)).

Gð3s, zÞ ¼ 1þgð3s, zÞ

ln zz0

� � ð6Þ

It is evident from Eqs. (4)–(6) that gust and peak factors should varywith the roughness length, z0, and/or the turbulence intensity, TIu.

Average gust and peak factors for each partition are presentedin Table 11. The gust factors are calculated using Eq. (3), whereU(3 s,z) is the observed maximum 3 s gust for each 15 min recordand U(z), the corresponding hourly mean wind, is obtained bydividing the 15 min mean by 1.05, in accordance with the open-terrain gust factor distribution from Durst (1960). Peak factors

Fig. 10. Gust factor distribution with corresponding turbulence intensity for those segm

and G(3, 10)¼1.52).

Fig. 11. Gust factor distribution with corresponding turbulence intensity for R

were derived by solving Eq. (5) and using the observed turbulenceintensity from the corresponding records. This approach will yielda conservative (higher) gust factor estimate since the use of a gustfactor curve implies that the 15 min average is the peak suchaverage within the hourly record. The Krayer and Marshall (1992)3-s gust factor, 1.65, is higher than the average gust factorcalculated for the partitions in Group 1, 1.524. However, the lattercompares well to the Durst (1960) value of 1.51 and to the ESDUgust factor, 1.53, obtained by Vickery and Skerjl (2005) afteranalyzing tropical cyclone data for relatively smooth terrain.

The availability of individual gust factors under strong winds isof crucial importance in numerical or physical wind engineering,as these practices deal with and assess wind loads on buildingsand structures. As an effort to satisfy this necessity, gust factorswere obtained for the 15 min segments with maximum windspeed at the 10 m level for each tower. These gust factors, whenplotted against the corresponding turbulence intensity (Fig. 10),exhibit a near linear variation with su/U.

Analysis of the results depicted in Fig. 10 shows that aturbulence intensity of 17.7% corresponds to the generally

ents with maximum wind speed (open exposure conditions correspond to TIu¼20%

ita T3 (open exposure conditions correspond to TIu¼20% and GF¼1.52).


accepted gust factor value of 1.52. Neutral atmospheric flowswith higher turbulence intensities are expected to have highergust factor values; however, for flows having lower turbulenceintensities lower gust factor values can be anticipated.Similar results for the entire data set from Rita T3 (Fig. 11)reveal a turbulence intensity of 16.5% for a 3 s gust factor of 1.52.The results from this study, which concur with the conclusiondrawn by Vickery and Skerjl (2005) after analyzing land- andocean-based weather systems, indicate that hurricane gust factorsshow a strong similarity to those of extra-tropical systems(e.g., winter storms).

Furthermore, it would be very instructive to obtain gust factorsfor the entire time range from 1 s to 1 h from the hurricane data.For this purpose one needs to select a group of four consecutive

Fig. 12. Gust Factor distribution for Katrina T2 (fi

Fig. 13. Gust Factor distribution for Rita T3 (file

15 min records for which the individual means are nearly equal.Location of a 1 h duration sample that is stationary with respect tomean flow and turbulence for data recorded in a fast movinghurricane is a challenging task. From the available data, the threemost stationary 1-h records were selected to obtain gust factordistributions at near peak mean wind speeds. The gust factorswere calculated according to Eq. (7), where U(t) is the t-s gustspeed obtained with a moving average and U(T¼3600 s) is theaverage wind speed for the 1 h record:

Gðt, T ¼ 3600sÞ ¼UðtÞ

UðT ¼ 3600sÞð7Þ

The gust factor distributions are shown in Figs. 12–14 forKatrina T2 and Rita T3. As observed previously, the distributions

les 36–39), U¼27.37 m/s, su/U¼16.9% (UTC).

s 24–27), U¼22.73 m/s, su/U¼26.8% (UTC).

Fig. 14. Gust Factor distribution for Katrina T2 (files 42–45), U¼26.49 m/s, su/U¼14.8% (UTC).


exhibit appreciable variation with turbulence intensity for shortaveraging times. The gust factor distribution for the data samplehaving a turbulence intensity of 16.9% matches the ASCE curvequite well (Fig. 12). The gust factor distribution from the samplewith a turbulence intensity of 26.8% exceeds the ASCE curveconsiderably for low averaging times (Fig. 13). On the other hand,for the low turbulence-intensity sample (14.8%) the observed gustfactors generally fall below the ASCE curve (Fig. 14).

11. Conclusions

Analyses of the ground winds observed during the passage ofhurricanes reveal a considerable amount of similarity to extra-tropical strong winds. The mean winds show the passage of thehurricane eye with two velocity peaks and the associated near1801 change in wind direction. Nevertheless, some of the mean-wind records exhibit a single velocity peak together with the 1801wind direction change. The observed turbulence intensities andthe turbulence derived aerodynamic roughness are associatedwith upwind terrain characteristics. Values of the observed uw

covariance are mostly negative as one would expect in a standardboundary layer. The 3 s gust factor increases with increasingturbulence intensities and its value of 1.52 corresponds to anatmospheric flow with a turbulence intensity of approximately17%.

Acknowledgements

The authors wish to thank the State of Florida Department ofCommunity Affairs Residential Construction Mitigation Programand the NOAA Florida Hurricane Alliance for supporting hurricanedeployment activities. Special recognition should be extended toDr. Kurtis Gurley (University of Florida) and Dr. Timothy Reinhold(Institute for Business and Home Safety) for their pioneeringefforts to establish the first in-situ wind engineering experimentsduring hurricane landfall. The authors also wish to thank Dr. Luis

Aponte for his assistance in figure preparation. This work wassupported in part by the National Science Foundation under Grantno. CMMI-0928563. Any opinions, findings, and conclusions orrecommendations expressed in this material are those of theauthors and do not necessarily reflect the views of the NationalScience Foundation.

References

ASCE 7-10, 2010. Minimum Design Loads for Buildings and Other Structures.American Society of Civil Engineers, Reston, VA.

Bradbury, W.M.S., Deaves, D.M., Hunt, J.C.R., Kershaw, R., Nakamura, K., Hardman,M.E., Hardman, P.W., 1994. The importance of convective gusts. Meteorolo-gical Applications 1, 365–378.

Counihan, J., 1975. Adiabatic atmospheric boundary layers—review and analysisof data from period 1880–1972. Atmospheric Environment 9 (10),871–905.

Durst, C.S., 1960. Wind speeds over short periods of time. Meteorological Magazine89, 181–1287.

Haan, F.L., Sarkar, P.P., Spencer-Berger, N.J., 2006. Development of an active gustgeneration mechanism on a wind tunnel for wind engineering and industrialaerodynamics applications. Wind and Structures 9 (5), 369–386.

Hogstrom, U., 1990. Analysis of turbulence structure in the surface layer witha modified similarity formulation for near neutral conditions. Journal ofAtmospheric Sciences 47, 55–78.

Izumi, Y., 1971, Kansas 1968 Field Program Data Report. Bedford, MA, Air ForceCambridge Research Papers, No. 379, 79pp.

Krayer, W., Marshall, R., 1992. Gust factors applied to hurricane winds. Bulletin ofthe American Meteorological Society 73 (5), 613–617.

Panofsky, H., Dutton, J., 1984. Atmospheric Turbulence, Models and Methods forEngineering Applications. Wiley International, NY, USA.

Salzano, C., Masters, F., Katsaros, J., 2010. Water penetration resistance ofresidential window installation options in hurricane prone areas. Buildingand Environment, 45 (6), 1373–1388.

Schroeder, J., Smith, D., 2003. Hurricane Bonnie wind flow characteristicsas determined from WEMITE. Journal of Wind Engineering and IndustrialAerodynamics 91, 767–789.

Schroeder, J., Edwards, B., Giammanco, I., 2009. Observed tropical cyclone windflow characteristics. Wind and Structures 12 (4), 347–379.

Vickery, P., Skerjl, P., 2005. Hurricane gust factors revisited. Journal of StructuralEngineering 131 (5), 825–832.

Wakimoto, R.M., Black, P.G., 1994. Damage survey of Hurricane Andrew and itsrelationship to the eyewall. Bulletin of the American Meteorological Society 75(2), 189–200.

Weber, R.O., 1999. Remarks on the definition and estimation of friction velocity.Boundary-Layer Meteorology 93, 197–209.


Wieringa, J., 1993. Representative roughness parameters for homogeneous terrain.Boundary-Layer Meteorology 63, 323–363.

Wurman, J., Winslow, J., 1998. Intense sub-kilometer-scale boundary layer rollsobserved in Hurricane Fran. Science 280, 555–557.

Yu, B., Chowdhury, A., Masters, F., 2008. Hurricane power spectra, co-spectra, andintegral length scales. Boundary-Layer Meteorology 129 (3), 411–430.

Zhou, Y., Kareem, A., 2002. Definition of wind profiles in ASCE 7. Journal ofStructural Engineering 128 (8), 1082–1086.

surface wind measurements in three gulf coast hurricanes ......surface wind measurements in three...

Documents